Last devlog before shipping:

Against my best wishes, none of my models reached a usable state. Even though their best loss is quite low (2.94 and 2.81) I didn’t manage to reach the finetuning stage.

To continue training I would have to start pouring money to get more compute power. And I don’t have that….

So that’ll be all.

I will finish the UI and deploy locally first until I manage to find a vps before my project get a reviewer.